The aim of this model is to increase the understanding of longer and more complex prompts.This model is highly experimental.<h1 id="recommended-weight-of-~0.75-az141wco9">Recommended weight of ~0.75</h1><h3 id="training-notes-lop7jx8kw">Training Notes</h3>The model is trained with very long and complex image captions. The intent was to see if by training with very detailed descriptions, I could increase the prompt understanding.Sometimes I feel like it's really increasing the prompt understanding, but it may certainly be bias. A lot more testing and training is needed!The dataset is now available, as well as a brief article on the creating it:<a target="_blank" rel="ugc" href="https://civarchive.com/articles/14046/the-caption-process-of-daiversity-detailer">https://civarchive.com/articles/14046/the-caption-process-of-daiversity-detailer</a>

dAIversity - Prompt Understanding Enhancer - XL Realistic (experimental)

dAIversityUnderstandingXLRealistic.safetensors

dAIversity Detailer LoRA Training Data v1.0.zip

v1.0

The purpose of this model is to add an increased range of diversity to your basic generations.When prompting for a person without a specific ethnicity specified, this model will make the outputs more diverse. It works very well at a distance, but has some issues with close-ups of people. To be improved in the future.Note:The diversity is not limited to people. The model is also trained on more diverse locations. A livingroom is not the same everywhere in the world.Hopefully this model unlocks and makes you things you haven't seen or experienced before.<h3 id="generation-settings-3x72v5gm6">Generation Settings</h3>It is flux, anything seems to work fine for me. I have no particular recommendations to give for now.

dAIversity Flux [Diverse checkpoint]

daiversityFluxDiverse_v10Fp8Checkpoint.safetensors

daiversityFluxDiverse_v10Fp8Checkpoint_trainingData.zip

v1.0-fp8-checkpoint

A detailer model that aims at adding details to your generations with Flux.The focus of the model is to enhance the details of photorealistic images.When used with other styles, it has an affect on the style of them, pulling them a bit towards realism. Use a lower weight to reduce the effect.<h1 id="recommended-weight-of-1-z7936z3kx">Recommended weight of 1</h1>The model is trained with very long and complex image captions mixed with keywords. It works equally well on simple and complex prompts.

Files with SHA256: fa15fb5ef9c70865177bc43467ab70c0ce88fc463df34b0b23e4cb4e3a083b58